Tetraspanin CD151 Regulates Glycosylation of α3β1 Integrin*

The tetraspanin CD151 forms a stoichiometric complex with integrin α3β1 and regulates its endocytosis. We observed that down-regulation of CD151 in various epithelial cell lines changed glycosylation of α3β1. In contrast, glycosylation of other transmembrane proteins, including those associated with CD151 (e.g. α6β1, CD82, CD63, and emmprin/CD147) was not affected. The detailed analysis has shown that depletion of CD151 resulted in the reduction of Fucα1–2Gal and bisecting GlcNAc-β(1→4) linkage on N-glycans of the α3 integrin subunit. The modulatory activity of CD151 toward α3β1 was specific, because stable knockdown of three other tetraspanins (i.e. CD9, CD63, and CD81) did not affect glycosylation of the integrin. Analysis of α3 glycosylation in CD151-depleted breast cancer cells with reconstituted expression of various CD151 mutants has shown that a direct contact with integrin is required but not sufficient for the modulatory activity of the tetraspanin toward α3β1. We also found that glycosylation of CD151 is also critical; Asn159 → Gln mutation in the large extracellular loop did not affect interactions of CD151 with other tetraspanins or α3β1 but negated its modulatory function. Changes in the glycosylation pattern of α3β1 observed in CD151-depleted cells correlated with a dramatic decrease in cell migration toward laminin-332. Migration toward fibronectin or static adhesion of cells to extracellular matrix ligands was not affected. Importantly, reconstituted expression of the wild-type CD151 but not glycosylation-deficient mutant restored the migratory potential of the cells. These results demonstrate that CD151 plays an important role in post-translation modification of α3β1 integrin and strongly suggest that changes in integrin glycosylation are critical for the promigratory activity of this tetraspanin.

Transmembrane proteins from the tetraspanin superfamily are assembled in microdomains (referred to as tetraspanin-en-riched microdomains (TERM) 2 ) which also incorporate a number of tetraspanin-interacting receptors (e.g. integrins and receptor tyrosine kinases) (1). It has been shown that tetraspanins regulate the activity of the associated receptors via various mechanisms involving ligand binding, clustering, and trafficking (1)(2)(3).
N-Linked glycosylation is one of the most common and diverse modifications of transmembrane proteins. The role of glycosylation in regulation of protein stability, folding, and dimerization and trafficking to and from the plasma membrane as well as between intracellular organelles is documented in numerous reports (4 -8). Furthermore, earlier studies have shown that glycosylation of a number of transmembrane proteins, which would be later identified as tetraspanin-associated partners, regulate their functions. Glycosylation of intercellular adhesion molecule 1 and components of major histocompatibility class I complex is important for interactions of the proteins with their respective receptors (9,10). Trafficking of CD4 to the cell surface was impaired in cells treated with tunicamycin (11). More recently, it was found that glycosylation of H,K-ATPase ␤ subunit, a partner for tetraspanin CD63, regulates internalization and subsequent degradation of the protein (12). A number of reports have described that surface expression, conformation, ligand binding, dimerization, and endocytosis of epidermal growth factor receptor are regulated by glycosylation (13)(14)(15). The N-linked glycan on ErbB3 prevents spontaneous heterodimerization and activation of the receptor (16).
There is a substantial body of evidence indicating that glycosylation of integrins plays a critical role in their function. Early experiments have shown that differential glycosylation of ␤2 integrin subunit is critical for its pairing with either ␣L or ␣M (17). Altered glycosylation of ␤1 integrins correlated with differences in attachment of cells to fibronectin and laminin (18). Similarly, binding of ␣5␤1 integrin to its ligand was decreased in cells treated with 1-deoxymannojirimycin, a compound that inhibits conversion of the high mannose to hybrid and complex glycosylated species of the protein (19). Subsequent work from various laboratories has extended these observations to other integrins, including ␣6␤1 (20), ␣v integrins (21), and ␣3␤1 (22)(23)(24)(25). In addition, it has been proposed that glycosylation-de-pendent interactions of integrins with their ligands involve gangliosides (26,27).
Recent studies have demonstrated that tetraspanin CD81 regulates glycosylation of its proximal partners CD19 and EWI-2 (30,31). In the case of CD19, the region responsible for this activity was mapped to the predicted N-terminal cytoplasmic portion of the protein (30). In addition, modulation in the expression levels of CD82/KAI-1 affected maturation and surface expression of ␤1 integrin subunit in lung carcinoma cells (32).
␣ chains of canonical laminin-binding integrins (i.e. ␣3␤1, ␣6␤1/␤4, and ␣7␤1) are post-translationally cleaved in Golgi by proprotein convertases to form light and heavy chains that are held together by a single disulfide bond and ␤ subunit (33). It is thought that cleavage of ␣ subunits is required for "inside-out" activation of integrin heterodimers (34). There are two splice variants of the light chains described for ␣3 subunit, each having three potential N-linked glycosylation sites (35).

Flow Cytometry and Cell Sorting
Semiconfluent cells were detached using Cell Dissociation Buffer (Invitrogen), incubated with saturating concentrations of primary mouse mAb for 1 h on ice, washed twice in phosphate-buffered saline, and then labeled with fluorescein isothiocyanate-conjugated goat anti-mouse IgG for 1 h at 4°C. Surface labeling was analyzed by flow cytometry using COULTER Epics XL. For sorting, cells were prepared as above except that all the solutions were sterilized by filtration.

Western Blot Analysis
Cells were lysed overnight at 4°C in 1% Triton X-100/ phosphate-buffered saline buffer containing inhibitors (2 mM phenylmethylsulfonyl fluoride, 10 g/ml aprotinin, 10 g/ml leupeptin). The lysate was centrifuged at 12,000 ϫ g for 10 min to remove insoluble material. Equal amounts of protein lysates were resolved by 12% SDS-PAGE under reducing and nonreducing conditions, transferred onto nitrocellulose membrane, and incubated with appropriate primary Ab. Protein bands were visualized after subsequent incubations with appropriate horseradish peroxidase-conjugated secondary Ab and Chemiluminescence Reagent Plus (PerkinElmer Life Sciences).

Immunoprecipitation
Cell lysates were prepared as above (except when the analysis of tetraspanin-tetraspanin association was performed) and precleared by incubation for 3 h at 4°C with agarose beads conjugated with goat anti-mouse antibodies (mIgGbeads; Sigma). Immune complexes were collected using appropriate mAbs prebound to the mIgG-beads and washed four times with the immunoprecipitation buffer. The complexes were eluted from the beads with Laemmli sample buffer. Proteins were resolved by SDS-PAGE, transferred to the nitrocellulose membrane, and developed with the appropriate Ab. For the analysis of ␣3␤1-tetraspanin (other than CD151) and tetraspanin-tetraspanin interactions, cell lysis and immunoprecipitation was carried out in 1% Brij 96 in the presence of aforementioned inhibitors.

Migration Assay
Migration was analyzed using a standard Boyden Chamber protocol. In brief, 1-2 ϫ 10 5 cells were detached using Cell Dissociation Solution (Invitrogen) and suspended in 500 l of serum-free DMEM. Cells were subsequently added into the inner compartment of Nunc's tissue culture inserts with polycarbonate membranes (8-m pores), the bottom sides of which were coated with 2 g/ml Laminin-332 or 10 g/ml fibronectin (Sigma). Cells were allowed to migrate toward serum-free DMEM supplemented with 10 ng/ml epidermal growth factor for 8 h. Nonmigrated cells were removed, and nuclei of migrating cells were stained with 4Ј,6-diamidino-2-phenylindole. Membranes were mounted on glass slides and analyzed using a Nikon Eclipse E600 microscope. Up to seven random fields per membrane were photographed and scored using ImageJ nuclear/cell counter program. Each of the experiments was done in quadruplicates, and 2-3 independent experiments were carried out for each cell line.

Deglycosylation Treatment
Protein lysate (ϳ100 g) was boiled for 10 min in the presence of 0.5% SDS before deglycosylation. The denatured lysate or immunoprecipitated complexes eluted in 0.2ϫ Laemmli buffer were deglycosylated overnight at 37°C with either peptide:N-glycanase F in Buffer G7 or endoglycosidase H in Buffer G5 (New England Biolabs). Samples were then resolved by 12% SDS-PAGE, transferred onto nitrocellulose membrane, and detected with Abs for various proteins. Deglycosylation was performed in both reducing (in the presence of 40 mM dithiothreitol) and nonreducing conditions.

CD151 Regulates Glycosylation of ␣3␤1
Integrin-The tetraspanin CD151 forms stoichiometric complexes with various laminin-binding integrins, including ␣3␤1, ␣6␤1, ␣6␤4, and ␣7␤1. To investigate whether CD151 affects biosynthetic processing of integrins, we established two epithelial cell lines in which expression of the tetraspanin was decreased, using a specific short hairpin RNA construct (45); MDA-MB-231/CD151(Ϫ) and HeLa/CD151(Ϫ) lines express CD151 at levels of Ͻ5% of the corresponding parental cells (Fig. 1A). Flow cytometry and Western blotting have shown that down-regulation of CD151 did not affect surface and total levels of ␣3␤1 and ␣6 integrins (␣7␤1 is not expressed in epithelial cells). Interestingly, we consistently observed that the protein band corresponding to the ␣3 integrin subunit runs slower in SDS-PAGE. This was particularly evident when we compared the positions of the ␣3 light chains. Light chains of ␣3 resolved as closely spaced three bands with the top, slower migrating band being more prominent in MDA-MB-231/CD151(Ϫ) than MDA-MB-231/ CD151(ϩ) cells (Fig. 1B). Conversely, the intensity of the lowest band was higher in the CD151-positive cells. The observed differences can be explained by one of the following: 1) differences in the furin-dependent cleavage; 2) differences in splicing; 3) differences in glycosylation. To distinguish between these possibilities, we compared the mobility of ␣3 light chains in SDS-PAGE after the treatment of protein lysates with peptide:N-glycanase. Since peptide:N-glycanase completely removes N-linked glycans from the protein backbone, this treatment would negate both qualitative and quantitative differences in glycosylation of the integrin in CD151-positive and CD151-negative cells. Indeed, we observed comparable patterns of the ␣3 light chains in the peptide:N-glycanase-treated lysates from the MDA-MB-231 and HeLa pairs ( Fig. 2A, lanes 3 and 4 and lanes 9 and 10). To examine differences in the integrin glycosylation in more detail, we analyzed mobility of the ␣3 light chains after treatment with endoglycosidase H, which specifically cleaves high mannose and hybrid forms of N-linked glycans. Although all three major glycoforms of the ␣3 light chain were sensitive to the endoglycosidase H treatment, the products of the digestion run slower in SDS-PAGE than completely deglycosylated species (Fig. 2A). These results indicated that the light chain of the ␣3 integrin subunit, which has three putative glycosylation sites, is modified by both complex and hybrid/high mannose oligosaccharides. Importantly, depletion of CD151 changed relative abundance of various endoglycosidase H-resistant glycoforms in MDA-MB-231 and HeLa cells ( Fig. 2A, lanes 5 and 6 and lanes 7 and  8). To gain further insight into the role of CD151 in matura-tion of ␣3␤1, we analyzed glycosylation of the integrin in cells treated with swainsonine and deoxymannojirimycin. These chemicals block activities of trimming glycosidases and, thereby, inhibit processing of N-linked oligosaccharides to complex (swainsonine) and hybrid/complex forms (deoxymannojirimycin). Although less pronounced, differences between CD151(ϩ) and CD151(Ϫ) cells were still visible after treatment with either swainsonine or deoxymannojirimycin (Fig. 2B). Taken together, these results demonstrated that CD151 influences glycosylation of ␣3␤1 at the relatively early stages of the maturation process.
Depletion of CD151 Does Not Affect Glycosylation of Other Glycoproteins within TERM; Association of Tetraspanins with ␣3␤1 Glycoforms-There are a large number of glycosylated proteins that are incorporated into TERM (63). To establish whether CD151 also targets other TERM-associated glycoproteins, we compared glycosylation patterns of CD63, CD82, and ␣6 integrin subunit. In addition, we analyzed glycosylation of EMMPRIN, a transmembrane protein that is known to associate with ␣3␤1 (51). The results of these experiments showed that glycosylation of none of these proteins was affected by depletion of CD151 ( Fig. 3A) (data not shown). It has been proposed that there exists a certain hierarchy of protein-protein interactions within TERM (64). Thus, we examined whether various ␣3␤1 glycoforms preferentially associate with other tetraspanins. As shown in Fig.  3B, there was a subtle difference in the patterns of ␣3 glycoforms co-precipitated by the mAbs to CD9 and CD81; although one of the two lower glycoform species predominated over the other in the CD81 immunoprecipitate, relative abundance of these forms in the CD9 immunoprecipitate was comparable (lanes 2 and 3). These differences in glycosylation between the CD9-and CD81-associated ␣3 were observed in three independent experiments.
To examine whether these tetraspanins can modify glycosylation pattern of ␣3␤1, we generated stable MDA-MB-231/CD9(Ϫ), MDA-MB-231/CD81(Ϫ) cell lines, in which expression of CD9 and CD81 was decreased by Ͼ95% (Fig. 3C). We also established a cell line (MDA-MB-231/ CD63(Ϫ)) depleted of tetraspanin CD63 (Fig. 3C). Immunoprecipitation and Western blotting experiments have shown  that down-regulation of these tetraspanins in MDA-MB-231 cells did not change mobility of the ␣3 subunits in SDS-PAGE (Fig. 3D). Collectively, our data show that 1) the effect of CD151 on glycosylation of ␣3␤1 integrin is specific, and 2) three other tetraspanins that are widely expressed in epithelial cells have no obvious role in glycosylation of this integrin.
The Role of Palmitoylation, Glycosylation, and the Large Extracellular Loop in the Activity of CD151 toward ␣3␤1-A recent study has established that glycosylation of CD19 is controlled by the N-terminal cytoplasmic domain of the tet-raspanin CD81 (30). To identify the region(s) required for the activity of CD151 toward ␣3␤1, we used MDA-MB-231/CD151(Ϫ) cells to generate a panel of stable lines that reexpressed various mutants of CD151 (the MDA-MB-231/rec series) (Fig. 4A). The following mutants were used in these experiments. 1) CD151palm(Ϫ) (palmitoylation-deficient CD151) was used. Previously, we have shown that in rat fibroblasts, this mutant can form direct contact with ␣3␤1, but recruitment of the complex into tetraspanin-enriched microdomains was impaired (38). 2) CD151glyco(Ϫ) (glycosylation-deficient CD151) was used. 3) CD151/ QRD has been previously described by Kazarov et al. (39), who showed that mutation of QRD to INF (residues 194 -196) within the large extracellular loop prevents direct interactions between CD151 and integrins when proteins are overexpressed in COS cells. Interactions of the CD151QRD mutant with other tetraspanins have not been affected. 4) CD151⌬C and SW6 mutants were described in our earlier studies. Both mutants retain their ability to associate with other tetraspanins and integrins (43,62). 5) We used 63-N-151, a chimeric mutant in which the N-terminal cytoplasmic region of CD151 was substituted for a corresponding part of CD63. Where appropriate, an additional mutation was introduced into the mutant DNA constructs to make them resistant to short hairpin RNA-CD151. We also reconstituted the expression of the wild-type CD151 in these cells to establish MDA-MB-231/rec/wt line, which was used as a positive control in these experiments. Western blotting showed that the expression levels of all mutants and the wild-type CD151 were comparable in the reconstituted cells (Fig. 4D, bottom). Two of the CD151 mutants used in this study (i.e. CD151glyco(Ϫ) and 63-N-151) have not been characterized previously. Hence, we analyzed whether these mutations affected interactions of the CD151 with other tetraspanins and ␣3␤1. As expected, both mutant proteins can be co-immunoprecipitated with CD9 and CD81 under the conditions that preserve tetraspanin-tetraspanin complexes (Fig. 4B). Furthermore, we found that ␣3␤1-CD151glyco(Ϫ) FIGURE 3. A, depletion of CD151 does not affect glycosylation of other transmembrane proteins. Cells were lysed in Triton X-100, and the lysates (or protein complexes immunoprecipitated with the mAb A6-ELE against ␣6 integrin subunit) were either left untreated or deglycosylated overnight with endoglycosidase H (lanes 3 and 4). Samples were resolved by 12% SDS-PAGE, transferred onto nitrocellulose membrane, and detected with rabbit Ab to ␣6 integrin subunit (top), mouse anti-CD63 mAb (1B5; middle), or mouse anti-emmprin mAb (8G6; bottom). B, association of CD9, CD81, and CD151 with glycoforms of ␣3␤1. MDA-MB-231 cells were lysed in Brij98/Triton X-100, and tetraspanin complexes were immunoprecipitated (IP) with anti-CD151, anti-CD9, or anti-CD81 mAbs. Immunoprecipitated protein complexes were eluted and deglycosylated overnight with endoglycosidase H (EndoH). Samples reduced with dithiothreitol were resolved by 12% SDS-PAGE, transferred onto nitrocellulose membrane, and detected with rabbit Ab to the ␣3 integrin subunit. Note the differences between CD9-and CD81-associated ␣3 with regard to relative abundance of lower molecular mass glycoforms (indicated with arrows). Similar differences were observed in three experiments. C, characterization of MDA-MB-231/CD63(Ϫ), MDA-MB-231/CD81(Ϫ), and MDA-MB-231/CD9(Ϫ) cells. Lysates of the cells were resolved by 12% SDS-PAGE and analyzed by Western blotting (WB) with mAbs against tetraspanin or ␤-actin. D, depletion of CD63, CD9, or CD81 does not affect glycosylation of ␣3 integrin subunit by MDA-MB-231 cells. Cells were lysed in Triton X-100, and ␣3␤1 integrin was immunoprecipitated with anti-␣3 mAb A3-IVA5. Immunoprecipitated protein complexes were eluted and deglycosylated overnight with endoglycosidase H. Samples reduced with dithiothreitol were resolved by 12% SDS-PAGE, transferred onto nitrocellulose membrane, and detected with rabbit Ab to ␣3 integrin subunit. DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51

JOURNAL OF BIOLOGICAL CHEMISTRY 35449
and ␣3␤1⅐63-N-151 complexes were resistant to Triton X-100 (Fig. 4C), thus suggesting that both mutants are in direct contact with ␣3␤1. Our experiments also confirmed that the Gln 194 -Arg 195 -Asp 196 sequence (QRD mutant) plays a critical role in the interaction of CD151 with ␣3␤1 in MDA-MB-231 cells (Fig. 4C, lane 5). We then analyzed the glycosylation pattern of ␣3 integrin subunit in MDA-MB-231/rec cells expressing CD151 mutants. As expected, we observed reconstitution of the glycosylation pattern for the ␣3 integrin subunit in MDA-MB-231/rec/wt cells (Fig. 4D,  compare lanes 1 and 3). Of all of the mutants examined, only CD151⌬C and 63-N-151 were able to restore fully the SDS-PAGE mobility pattern of the ␣3 light chain (Fig. 4D, lanes 6 and 9). Based on these results, we have drawn the following conclusions. First, the presence of the intact CD151 large extracellular loop (and, therefore, direct contact between CD151 and ␣3␤1) is not sufficient for the ability of the tetraspanin to modulate glycosylation of the integrin (results with MDA-MB-231/ rec/SW6 and MDA-MB-231/rec/ palm(Ϫ) cells; Fig. 4D, lanes 4 and  8). Second, neither N-nor C-terminal cytoplasmic regions are required for this purpose (Fig. 4D, lanes 6  and 9). Third, palmitoylationdependent compartmentalization of CD151 to tetraspanin-enriched microdomains is essential for the activity of the protein toward ␣3␤1 (results with MDA-MB-231/rec/ palm(Ϫ) cells; Fig. 4D, lane 4).  3 and 4  and lanes 7 and 8). These data indicated that depletion of CD151 resulted in a decrease in bisecting GlcNAc residues and Fuc␣1-2Gal N-glycans. By contrast, binding of P. vulgaris leucoagglutinin, which preferentially reacts with ␤1,6-branched GlcNAc, and wheat germ agglutinin, which binds to GlcNAc and sialic acid, was comparable in CD151-positive and CD151depleted cells (Fig. 5, lanes 5 and 6 and lanes 9 and 10).
Changes in Glycosylation in CD151-depleted Cells Correlate with Inhibition of ␣3␤1-dependent Cell Migration-To investigate the functional consequences of CD151-dependent changes in glycosylation of ␣3␤1 in MDA-MB-231 cells, we compared migration of CD151-positive and CD151-de- pleted cells toward laminin-332. The initial antibody-blocking experiments have shown that migration of MDA-MB-231 cells toward laminin-332 requires both ␣3␤1 and ␣6 integrins. The inhibitory effect was more dramatic when cells were pretreated with the anti-␣3 mAb (ϳ90% of inhibition; Fig. 6A), compared with the inhibitory anti-␣6 mAb, which reduced migration by ϳ40%. The major contribution of ␣3␤1 integrin to migration of MDA-MB-231 cells toward laminin-332 was further confirmed when we analyzed motility of the ␣3-depleted cells (results are not shown). We then compared migration of MDA-MB-231/CD151(Ϫ) and MDA-MB-231/CD151(ϩ) cells in Boyden chamber assays. As illustrated in Fig. 6B, depletion of CD151 dramatically reduced the migratory potential of MDA-MB-231 toward laminin-332. On the other hand, migration toward fibronectin was unaffected (Fig. 6C). Importantly, attachment of MDA-MB-231 to laminin-332 or fibronectin (as examined in static adhesion assays (30 -40 min)) was not affected by the depletion of CD151 (Fig. 6D). These results provided further support for a specific functional link between CD151 and laminin-binding integrins during cell migration. We also analyzed migratory properties of MDA-MB-231/rec/ glyco(Ϫ) cells. As shown above, this point mutant of CD151 displayed biochemical properties similar to those of the wild-type protein (i.e. the ability to associate with other tetraspanins and form tight complexes with integrins) but could not restore the glycosylation pattern of the ␣3 integrin subunit. Therefore, the CD151glyco(Ϫ) mutant was particularly useful to analyze the impact of the CD151-dependent changes in integrin glycosylation on its function. Notably, although reexpression of the wild-type CD151 restored migratory potential of the cells, migration of cells expressing the glycosylation-deficient mutant remained at the level of CD151-deficient MDA-MB-231 (Fig. 6B). Collectively, these data strongly suggest that CD151dependent modulation of ␣3␤1 glycosylation is functionally relevant and plays an important regulatory role in cell migration toward laminin-332.

DISCUSSION
In this report, we identified tetraspanin CD151 as a modulator of glycosylation of ␣3␤1 integrin. Importantly, glycosylation of other proteins that are associated with tetraspanin microdomains was not affected.
Although tetraspanin-dependent modulation of protein glycosylation is not a unique feature of CD151 (see Introduction), the underlying mechanisms are likely to be different. Indeed, although the N-terminal cytoplasmic end is critical for the modulatory activity of CD81 (30), a corresponding region of CD151 is not required. The differences in glycosylation of the ␤1 inte-   DECEMBER 19, 2008 • VOLUME 283 • NUMBER 51 grin subunit observed in cells overexpressing CD82 are likely to be due to a more general role of this tetraspanin in biosynthetic trafficking and/or delivering of integrins to the cell surface (32). In this regard, in agreement with previous data (45,66), we found that CD151 does not affect surface levels of the associated integrins. 3 Our results show that CD151 affects glycosylation of ␣3␤1 integrin through multiple and interdependent mechanisms (Fig. 7). The most important factor seems to be the recruitment of ␣3␤1 to the tetraspanin-enriched microdomains. This process requires direct interactions of CD151 with the integrin on one side and with other tetraspanins on the other. When either of these interactions was affected (as for the QRD and palmitoylation mutants), the modulatory activity of CD151 toward ␣3␤1 was lost (Fig. 7). The modulatory deficiency of the CD151palm(Ϫ) mutant is particularly interesting. Previous data have shown that although early steps in the assembly of the ␣3␤1-CD151 complex are not affected by the mutations, recruitment of the complex to TERM, which is thought to take place in Golgi, is impaired (38,67). Thus, our data indicate that TERM can be used as a platform for diversification of the glycosylation process.

CD151 Regulates Glycosylation of ␣3␤1 Integrin
Although most tetraspanins possess one or more potential N-linked glycosylation sites, the role of glycosylation in various tetraspanin-associated activities remains largely unexplored. Studies involving CD82/KAI-1 have shown that glycosylation regulates interaction of the protein with integrins ␣5␤1 and ␣3␤1 and correlates with the antimigratory activity of the tetraspanin (68,69). Differences in glycosylation of tetraspanin CD63 were observed during maturation of dendritic cells. This correlated with the changes in the morphology of the major histocompatibility complex class II compartments that are thought to be responsible for presentation of major histocompatibility complex II-peptide complexes on the cell surface (70). Finally, specific glycosylation of uroplakin Ia/TSPAN-20 is critical for FimH lectin-dependent binding of Escherichia coli to the surface of urothelial cells (71). Our data indicate for the first time that glycosylation of a tetraspanin protein modulates glycosylation of its specific partner. The CD151glyc(Ϫ) mutant could not restore the glycosylation pattern of ␣3␤1 in MDA-MB-231/CD151(Ϫ) cells. Importantly, the interaction with ␣3␤1 and recruitment of the complex to the tetraspanin-enriched microdomains was not affected by this mutation. Hence, direct contact with ␣3␤1 and its compartmentalization to TERM are not the only factors that control the repertoire of oligosaccharides attached to the integrin complex. How do N-linked oligosaccharides on CD151 affect glycosylation of ␣3␤1? One possibility is that oligosaccharides attached to CD151 cause subtle changes in the conformation of ␣3␤1 complex that make integrin subunits a preferable substrate for particular glycosyltransferases. Alternatively, CD151 may function as a scaffolding platform for specific glycosyltransferases, which, while glycosylating the tetraspanin itself, may also act on the associated integrin complex (Fig. 7). In this regard, previous studies have shown the existence of a sequential glycosylation mechanism whereby a preceding glycosyltransferase activity determines the following glycosyltransferase that will be recruited for glycosylation. Thus, the presence of CD151 may initiate the recruitment of a particular glycosyltransferase (or a multienzyme complex) to ␣3␤1 and determine the order of subsequent glycosyltransferases that will come into contact with the integrin. Altered glycosylation of the SW6 mutant 3 G. Baldwin and F. Berditchevski, unpublished results.

FIGURE 7. Schematic diagram showing interrelationship between ␣3␤1
and CD151 during integrin maturation. The complex between CD151 and ␣3 integrin subunit is formed early during the biosynthesis. Association with CD151 is required for recruitment of ␣3␤1-CD151 to tetraspanin-enriched microdomains. CD151 regulates glycosylation of ␣3␤1 late in the endoplasmic reticulum (ER) and in Golgi. In the absence of CD151 (or when cells express palmitoylation-deficient CD151 and CD151 mutant, which lost direct contact with ␣3␤1), the integrin is not recruited to TERM. Consequently, fucosylation (i.e. Fuc␣1-2Gal) of ␣3␤1 and formation bisecting GlcNAc-␤(134) linkage are decreased. This may be due to the differences in the abundance of various glycosyltransferase complexes in TERM and non-TERM compartments. Although glycosylation-deficient CD151 (CD151glyco(Ϫ)) retains a direct contact with ␣3␤1 and targets the integrin to TERM, recruitment of specific GT complexes to the CD151-␣3␤1 complex may be affected, and this suppresses modifying activity of the tetraspanin.
(judged by the altered mobility in SDS-PAGE when compared with the wild-type protein) and, therefore, a distinct repertoire of recruited glycosyltransferases may explain inability of this mutant to restore the glycosylation pattern of ␣3␤1.
Specific glycomodifications of the integrin within TERM may involve a more favorable presentation of ␣3␤1 to (or shielding from) a TERM-associated glycosyltransferase. We found that depletion of CD151 specifically reduces Fuc␣1-2Gal and bisecting GlcNAc residues on the N-glycan of ␣3 integrin subunit (as revealed by the decreased binding of lectins). It is currently unknown how activities of fucosyltransferases I and II (FUT1 and FUT2) and N-acetylglucosaminyltransferase III (GnT-III), enzymes responsible for Fuc␣1-2Gal and bisecting GlcNAc modifications, respectively, are regulated by CD151. Sasai et al. (72) have shown that GnT-III is associated with caveolin-1. Furthermore, these authors proposed that this interaction regulates localization of the enzyme within Golgi and, ultimately, affects glycosylation pattern of cellular proteins. Although CD151 is not directly associated with caveolin (73), it is possible that, like some of the other tetraspanins (e.g. CD82 (74)), it may change compartmentalization of glycosyltransferases through a more general effect on lipid composition of cellular membranes.
We found that a decrease in bisecting GlcNAc structures on the ␣3 integrin subunit correlated with inhibition of tumor cell migration (see also below). By contrast, Zhao et al. (23) have recently reported that down-regulation of GnT-III and, consequently, decreased biosynthesis of bisecting GlcNAc residues resulted in an increased motility of neuroblastoma cells. There are at least two possible explanations for these seemingly contradictory results. First, Zhao et al. (23) found that in their cellular model, there was an inverse correlation between the amounts of bisecting GlcNAc and ␤1,6-GlcNAc-branched N-glycans on the ␣3 integrin subunit (23). Therefore, it is possible that in neuroblastoma cells, genuine suppression of cell motility associated with a decrease in bisecting N-glycans might have been negated by a strong promigratory activity of ␤1,6-GlcNAc. In this regard, we demonstrated that depletion of CD151 in MDA-MB-231 cells did not change the amount of ␤1,6-GlcNAc structures on the ␣3 subunit. Second, we found that the effect of CD151 on glycosylation of ␣3 integrin subunit in MDA-MB-231 cells is not limited to a decrease in bisecting GlcNAc structures. Indeed, as discussed above, depletion of CD151 also decreased Fuc␣1-2Gal residues on the ␣3 integrin subunit. Thus, it is feasible that this (or any other CD151-dependent changes in integrin glycosylation) could reverse the anti-migratory activity of bisecting GlcNAc structures.
Our results point to a new mode of regulation of integrin function by tetraspanins. We and others have previously suggested that CD151 controls cell motility by modulating integrindependent signaling and/or through regulation of integrin endocytosis and recycling (45,49,75). Here we established a strong correlation between the effect of CD151 on glycosylation of ␣3␤1 and promigratory activity of this tetraspanin. Indeed, the CD151glyco(Ϫ) mutant with only a single change of amino acid within the large extracellular loop was unable to restore the glycosylation pattern of ␣3 integrin subunit in MDA-MB-231/CD151(Ϫ) cells and ␣3␤1-dependent cell motility. Importantly, attachment of CD151-depleted MDA-MB-231 cells to laminin-332 in static adhesion experiments was not affected. Thus, the decrease in cell motility associated with CD151-dependent modifications in ␣3␤1 glycosylation may be due to a more subtle effect on interactions of the integrin with laminin-332 (e.g. transient interactions at the leading edge). Alternatively, changes in glycosylation may influence the rate of integrin turnover in migrating cells. This may involve gangliosides (76), whose interactions with integrins require N-linked glycans (65). Further work will be necessary to distinguish between these possibilities.